Paper Overview
Field: Computer Vision Authors: Md Thamed Bin Zaman Chowdhury, Moazzem Hossain Released: 2025-08-26 arXiv: 2508.17630
Summary
Road traffic injuries remain a major challenge in low- and middle-income countries, where proactive road safety auditing is limited by incomplete crash records, shortages of qualified auditors, and the high cost of large-scale field inspections.
This paper proposes Expert-Grounded Distillation (EGD), a novel AI framework that transfers institutional road safety expertise into a compact vision-language model for scalable visual road safety auditing. The key innovation is a quantified expert-grounding stage in which the teacher vision-language model is calibrated against authoritative field audits. Large-scale annotation is permitted only after the teacher reaches substantial agreement with expert risk assessments (Cohen's kappa = 0.74).
The calibrated teacher generates structured supervision signals that are distilled, via LoRA and a single leak-free prompt, into an 8-billion-parameter student model.
The authors also release BD-ARSA, the first open, expert-grounded visual road safety auditing dataset from Bangladesh, containing 21,947 image-audit records, along with the purpose-built EG-ARSA model.
Key Findings
- Grounded fine-tuning significantly outperforms zero-shot baselines on visual road safety auditing.
- The compact 8B-parameter student model surpasses both its 31B-parameter teacher and Gemini-2.5-Flash in blind evaluation.
- The calibration-before-annotation gate (Cohen's kappa = 0.74) ensures annotation quality before large-scale data generation.
- Paper: https://arxiv.org/abs/2508.17630